Methodology

Where our data comes from

Fixtures, scores, standings, line-ups and bookmaker odds come from a third-party data provider, API-Football. A background job (the sync worker) downloads this data on a schedule and stores it in our own database; pages read that database or our own API, not a call from your browser to the provider.

Scores of matches in play are polled about every minute. Pre-match odds are refreshed about every three hours, so a price on WBookie can be several hours older than the price at the bookmaker. The odds tools show the time the data was fetched.

The "opening" price we show is the first price WBookie saw for that match, which can be later than the bookmaker's real opening price.

Odds, implied probability and margin

Decimal odds convert to an implied probability with 1 ÷ odds. In a three-way market the implied probabilities add up to more than 100%; the excess is the bookmaker's margin. Example: odds of 2.00, 3.50 and 4.00 imply 50.0%, 28.6% and 25.0%, a total of 103.6%, so the margin is 3.6%.

To get a fair probability we remove the margin proportionally: each implied probability is divided by the total. In the example that gives 48.3%, 27.6% and 24.1%. This is the method behind the crowd-versus-bookmakers panel on match pages (the average price across bookmakers, margin removed) and behind the value-bet tool.

• Value bets: we compare each bookmaker's price with the fair price of one low-margin reference bookmaker, taken in the same data refresh. A price at least 1% above the fair price is listed; anything above 30% is discarded as a probable data error. A match without the reference bookmaker's prices has no value bets.
• Dropping odds: the percentage fall of a price against the first price we saw, or against the previous update. Falls of at least 5% are stored; the page filters from 10% by default.
• Arbitrage: we add 1 ÷ the best price of every outcome in a market, using prices from the same refresh. A total below 1 means a theoretical profit. Results above 20% are discarded as data errors. It is theory only: prices may have changed since the refresh.

How tipsters are ranked

A tip on WBookie is structured: one match, one market (1X2, double chance, over/under, Asian handicap or both teams to score), one selection and a virtual stake of 1 to 10 units. Our server sets the odds from the stored bookmaker prices when the tip is submitted; tipsters cannot type their own odds. Tips close 60 seconds before kick-off, each person gets one tip per market per match, and a submitted tip cannot be edited.

Grading is automatic. A job checks pending tips every ten minutes against the final score after 90 minutes (extra time and penalties do not count). A match that is cancelled, abandoned or never reaches a final state is voided after 72 hours.

Ranking uses ROI: profit divided by the amount staked, times 100, over won, lost and pushed (stake returned) tips. A tipster needs at least 10 settled bets to rank above those with fewer, so two lucky wins cannot top the table. Monthly and quarterly seasons use Vietnam time; a season ranks only tipsters with at least 10 settled tips (month) or 25 (quarter); ties are broken by profit and then by number of tips; a season is closed 76 hours after it ends so that every tip has been graded.

Simulated tipsters (bots) are labelled as bots and ranked separately from people. Only people earn badges. Stakes and winnings are virtual: there are no cash prizes, and WBookie takes no money from tipsters.

Predictions and match analysis

WBookie does not publish a proprietary forecasting model. What appears on match and prediction pages comes from three sources: the tips and votes of our tipsters and readers; the probabilities implied by bookmaker odds with the margin removed (see above); and our own match previews and reports, assembled from the fixture, standings and odds data in our database.

Every probability describes prices or opinions, not a promise. No prediction on WBookie is betting advice.

How articles are produced

Some articles, such as match previews, match reports and daily odds analysis, are generated from templates filled with the data described above. Parts of a preview, for example the team-news section, may be expanded by an AI language model. Today some of these generated articles are still published automatically. We are moving all of them to a draft workflow in which an editor reads each one and can change or reject it before it goes live; this page will say so once that is in place.

Each article shows its publication date and the date of its last update; the update date stored with the article is the one search engines read. When we correct an article, that date changes.

Limits and corrections

Data can arrive late, incomplete or wrong at the source. Odds and scores may differ from the bookmaker's or the competition's own site; check before you act.

How we correct errors is described in the editorial policy.

Prediction model (pdc-1)

WBookie is launching a statistical prediction model, pdc-1: a Dixon–Coles Poisson model fitted per league on recent results, with more weight on the most recent matches. It estimates each team's attacking and defensive strength and turns them into probabilities for 1X2, over/under 1.5, 2.5 and 3.5 goals, both teams to score and the correct score.

Each prediction is locked between 96 hours and 60 seconds before kick-off and is never changed afterwards. After the match it is graded on the 90-minute score with the hit rate, the Brier score, the ranked probability score and the log loss, next to a market baseline (average, margin removed) and a frequency baseline.

Every locked prediction stays public, misses and void ones included, and each carries a hash so the monthly CSV (CC BY 4.0) can be checked independently. A probability is an estimate, not a promise about a result.

Status: launching. Predictions are not published yet; this section will link the public ledger once it opens.